Data Driven Traffic Graph Construct Method

The electronic vector map is the core component of the intelligent transportation system,and it is the basis for realizing traffic flow forecasting,travel planning,traffic situation presentation,and macro and micro traffic management.The production of existing electronic maps is mainly realized along two technical routes.First,use geographic information software such as Arc GIS to label,organize,and draw remote sensing images into a map by human-computer interaction;Second,use vehicles equipped with high-precision GPS and radar equipment to scan,aggregate,and calculate the highway..However,these methods require a lot of manpower and material resources,and the production cycle is long and the dynamic update is slow.How to construct road vector maps more efficiently,quickly and accurately has always been a hot research issue in academia and business circles.This paper analyzes and studies the characteristics of GPS trajectory data and remote sensing image data,and comprehensively uses the two kinds of data to propose a set of electronic map drawing procedures.The work of this article includes the following three points:(1)For remote sensing image data,a road topology extraction method based on road tracking is proposed.The core idea is to use remote sensing images as input to predict the direction of road expansion at specific points(such as intersections)by a neural network model,and to make the road topology map grow in the predicted direction;on this basis,a complete Road topology graph is generated.In addition,this article introduces several road vector quality evaluation methods.Comparative experiments verify the effectiveness of the proposed method.(2)For GPS data,a method of using GPS angle information to describe road features is proposed.The core idea is to introduce a convolutional neural network to learn the mapping relationship between the road centerline and road features,and use post-processing methods to transform the network output into a topological structure map.Based on real GPS trajectory data,experimental analysis verifies the effectiveness of the proposed method.(3)A road level estimation method is designed.First,a road segmentation model that integrates multi-scale feature information is designed;then,on this basis,the direction template matching method use both the road topology map and the road segmentation results in the remote sensing image to estimate the level of the road.Experiments verify the effectiveness of the proposed method.